Course Overview Yu Hen Hu Introduction to ANN & Fuzzy Systems
Outline Overview of the course Goals, objectives Background knowledge required Course conduct Content Overview (highlight of each topics) 2
What is this Course Goal: Be familiar with modern advanced data processing tools that are motivated and originated from artificial neural network and fuzzy logic system fields. Understand basic theory and characteristics of these tools, and Occasions/applications that these tools may be applicable or may not be applicable Proper usage of these tools, expectation of the outcomes and interpretation of the results. Emphasis: Awareness Applications Implementations Many of these tools are readily available: Matlab neural network toolbox, fuzzy logic toolbox, and public domain implementations (see course home page) The instructor also provide sample implementation of many algorithms, as well as a number of public-domain implementations. 3
Textbook Neural Networks: A Comprehensive Foundation, Simon Haykin, Prentice Hall, New Jersey, second edition, 1999. (required) Neural Networks and Learning Machines Simon Haykin, Pearson Education Inc., Upper Saddle River, NJ 07458, third edition 2009. (Optional.) This is the third edition of the text book with more updated materials. This book can be used in lieu of the text book) Handbook of Neural Network Signal Processing, Y. H. Hu and J.- N. Hwang, CRC Press, 2002 ( Optional ) Advanced Fuzzy Systems Design and Applications, Yaochu Jin, Physica-Verlag Heidelberg, 2003, ISBN 3-7908-1537-3. (Optional) A set of class notes will be available on the web. Instructor will give password to access the notes. 4
Background Knowledge Required Calculus Familiar with derivatives, integration, Knowledge of vector calculus such as gradient Linear algebra: Familiar with matrices, vectors, inner product operations, Know what are matrix inversion, eigenvalues, singular values, subspace Probability and statistics: Probability, distribution, density function, Bayes rule Understand mean, variance, expectation, normal distribution Tutorials of these topics are available at course resource links. http://homepages.cae.wisc.edu/~ece539/resources/link.html 5
Programming Matlab will be used for all examples. Neural net toolbox and fuzzy logic toolbox are useful but not required. All Matlab m-files used in class will be posted in the course web page. http://homepages.cae.wisc.edu/~ece539/matlab/index.html Public domain software will be listed on course web page. http://homepages.cae.wisc.edu/~ece539/software/index.html These include both Matlab and C program implementation of various neural network paradigms. Projects may be conducted using Matlab, C or C++, or any programming languages. 6
Homework Course Conduct Three to five homework sets. Many problems will require programming or running ANN/fuzzy software to perform numerical experiments. One take home final examination Individual course project, Including project proposal, project report, and power point presentation. Electronic copies of these items will be posted on course web page. 3-5 min project presentation will be video taped and posted on line (most likely Youtube ) Grading 30% three to four assignments 30% Take home final exam 40% Individual Term Project 7
Tentative Topics To Be Covered ANN Basics, neurons, learning algorithms Perceptron learning, Pattern classification and function approximation Time series prediction, system identification Multi-Layer Perceptron (MLP), back-propagation learning, and applications Support vector machine (SVM) Radial Basis Network Fuzzy Set Theory and Fuzzy Logic Control Committee machine Genetic Algorithm and Evolution Computing Clustering, Self-Organization Map, Learn vector quantization Bayesian network and Hidden Markov Model (time permitting) 8